Group Task

Objectives and mandatory items

The objective of the delivery is to perform an analysis of the electoral data, carrying out the debugging, summaries and graphs you consider, both of their results and the accuracy of the electoral polls.

Specifically, you must work only in the time window that includes the elections from 2008 to the last elections of 2019.

General comments

In addition to what you see fit to execute, the following items are mandatory:

• Each group should present before 9th January (23:59) an analysis of the data in .qmd and .html format in Quarto slides mode, which will be the ones they will present on the day of the presentation.

• Quarto slides should be uploaded to Github (the link should be provided by a member of each group).

• The maximum number of slides should be 40. The maximum time for each group will be 20-22 minutes (+5 minutes for questions).

• During the presentation you will explain (summarised!) the analysis performed so that each team member speaks for a similar amount of time and each member can be asked about any of the steps. The grade does not have to be the same for all members.

• It will be valued not only the content but also the container (aesthetics).

• The objective is to demonstrate that the maximum knowledge of the course has been acquired: the more content of the syllabus is included, the better.

Mandatory items:

1. Data should be converted to tidydata where appropriate.

2. You should include at least one join between tables.

3. Reminder: information = variance, so remove columns that are not going to contribute anything.

4. The glue and lubridate packages should be used at some point, as well as the forcats. The use of ggplot2 will be highly valued.

5. The following should be used at least once:

   • mutate
   • summarise
   • group_by (or equivalent)
   • case_when

6. We have many, many parties running for election. We will only be interested in the following parties:

   • PARTIDO SOCIALISTA OBRERO ESPAÑOL (beware: it has/had federations - branches - with some other name).
   • PARTIDO POPULAR
   • CIUDADANOS (caution: has/had federations - branches - with some other name)
   • PARTIDO NACIONALISTA VASCO
   • BLOQUE NACIONALISTA GALLEGO
   • CONVERGÈNCIA I UNIÓ
   • UNIDAS PODEMOS - IU (beware that here they have had various names - IU, podem, ezker batua, …- and have not always gone together, but here we will analyze them together)
   • ESQUERRA REPUBLICANA DE CATALUNYA
   • EH - BILDU (are now a coalition of parties formed by Sortu, Eusko Alkartasuna, Aralar, Alternatiba)
   • MÁS PAÍS
   • VOX

7. Anything other than any of the above parties should be imputed as “OTHER”. Remember to add properly the data after the previous recoding.

8. Party acronyms will be used for the visualizations. The inclusion of graphics will be highly valued (see https://r-graph-gallery.com/).

9. You must use all 4 data files at some point.

10. You must define at least one (non-trivial) function of your own.

11. You will have to discard mandatory polls that:

-   refer to elections before 2008
-   that are exit polls
-   have a sample size of less than 750 or are unknown
-   that have less than 1 or less fieldwork days

12. You must obligatorily answer the following questions (plus those that you consider analyzing to distinguish yourself from the rest of the teams, either numerically and/or graphically)

-   Which party was the winner in the municipalities with more than 100,000 habitants (census) in each of the elections?
-   Which party was the second when the first was the PSOE? And when the first was the PP?
-   Who benefits from low turnout?
-   How to analyze the relationship between census and vote? Is it true that certain parties win in rural areas?
-   How to calibrate the error of the polls (remember that the polls are voting intentions at national level)?
-   Which polling houses got it right the most and which ones deviated the most from the results?

You should include at least 3 more “original” questions that you think that it could be interesting to be answer with the data.

Marks

The one who does the most things will not be valued the most. More is not always better. The originality (with respect to the rest of the works, for example in the analyzed or in the subject or …) of what has been proposed, in the handling of tables (or in the visualization), the caring put in the delivery (care in life is important) and the relevance of what has been done will be valued. Once you have the mandatory items with your database more or less completed, think before chopping code: what could be interesting? What do I need to get a summary both numerical and visual?

Remember that the real goal is to demonstrate a mastery of the tools seen throughout the course. And that happens not only by the quantity of them used but also by the quality when executing them.

Some dataviz will be extremely positive valued.

Required packages

Insert in the lower chunk the packages you will need

rm(list = ls())
library(tidyverse)

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream ======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

Warning: il pacchetto 'ggplot2' è stato creato con R versione 4.4.2

Warning: il pacchetto 'stringr' è stato creato con R versione 4.4.2

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr     1.1.4     ✔ readr     2.1.5
✔ forcats   1.0.0     ✔ stringr   1.5.1
✔ ggplot2   3.5.1     ✔ tibble    3.2.1
✔ lubridate 1.9.3     ✔ tidyr     1.3.1
✔ purrr     1.0.2     
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

library(dplyr)
library(tidyr)
library(stringr)
library(lubridate)
library(DataExplorer)

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

library(dplyr)
library(tidyr)
library(stringr)
library(lubridate)
library(DataExplorer)
library(glue)

Warning: il pacchetto 'glue' è stato creato con R versione 4.4.2

library(scales)

Warning: il pacchetto 'scales' è stato creato con R versione 4.4.2

Caricamento pacchetto: 'scales'

Il seguente oggetto è mascherato da 'package:purrr':

    discard

Il seguente oggetto è mascherato da 'package:readr':

    col_factor

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

Data

The practice will be based on the electoral data archives below, compiling data on elections to the Spanish Congress of Deputies from 2008 to the present, as well as surveys, municipalities codes and abbreviations.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

# NO TOQUES NADA
election_data <- read_csv(file = "./data/datos_elecciones_brutos.csv")

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# NO TOQUES NADA
election_data <- read_csv(file = "./data/datos_elecciones_brutos.csv")

Warning: One or more parsing issues, call `problems()` on your data frame for details,
e.g.:
  dat <- vroom(...)
  problems(dat)

>>>>>>> Stashed changes

Rows: 52206 Columns: 471
── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr  (11): tipo_eleccion, mes, codigo_ccaa, codigo_provincia, codigo_municip...
dbl (422): anno, vuelta, codigo_distrito_electoral, numero_mesas, censo, par...
lgl  (38): PARTIDO CARLISTA DE EUSKALHERRIA-EUSKALHERRIKO KARLISTA ALDERDIA,...

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

cod_mun <- read_csv(file = "./data/cod_mun.csv")

=======

cod_mun <- read_csv(file = "./data/cod_mun.csv")

>>>>>>> Stashed changes =======

cod_mun <- read_csv(file = "./data/cod_mun.csv")

>>>>>>> Stashed changes =======

cod_mun <- read_csv(file = "./data/cod_mun.csv")

>>>>>>> Stashed changes

Rows: 8135 Columns: 2
── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr (2): cod_mun, municipio

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

surveys <- read_csv(file = "./data/historical_surveys.csv")

=======

surveys <- read_csv(file = "./data/historical_surveys.csv")

>>>>>>> Stashed changes =======

surveys <- read_csv(file = "./data/historical_surveys.csv")

>>>>>>> Stashed changes =======

surveys <- read_csv(file = "./data/historical_surveys.csv")

>>>>>>> Stashed changes

Rows: 3753 Columns: 59
── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr   (4): type_survey, id_pollster, pollster, media
dbl  (51): size, turnout, UCD, PSOE, PCE, AP, CIU, PA, EAJ-PNV, HB, ERC, EE,...
lgl   (1): exit_poll
date  (3): date_elec, field_date_from, field_date_to

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

abbrev <- read_csv(file = "./data/siglas.csv")

=======

abbrev <- read_csv(file = "./data/siglas.csv")

>>>>>>> Stashed changes =======

abbrev <- read_csv(file = "./data/siglas.csv")

>>>>>>> Stashed changes =======

abbrev <- read_csv(file = "./data/siglas.csv")

>>>>>>> Stashed changes

Rows: 587 Columns: 2
── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr (2): denominacion, siglas

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

The data will be as follows:

• election_data: file with election data for Congress from 2018 to the last ones in 2019.

  • tipo_eleccion: type of election (02 if congressional election)
  • anno, mes: year and month of elections
  • vuelta: electoral round (1 if first round)
  • codigo_ccaa, codigo_provincia, codigo_municipio, codigo_distrito_electoral: code of the ccaa, province, municipality and electoral district.
  • numero_mesas: number of polling stations
  • censo: census
  • participacion_1, participacion_2: participation in the first preview (14:00) and second preview (18:00) before polls close (20:00)
  • votos_blancos: blank ballots
  • votos_candidaturas: party ballots
  • votos_nulos: null ballots
  • ballots for each party

• cod_mun: file with the codes and names of each municipality

• abbrev: acronyms and names associated with each party

• surveys: table of electoral polls since 1982. Some of the variables are the following:

  • type_survey: type of survey (national, regional, etc.)
  • date_elec: date of future elections
  • id_pollster, pollster, media: id and name of the polling company, as well as the media that commissioned it.
  • field_date_from, field_date_to: start and end date of fieldwork
  • exit_poll: whether it is an exit poll or not
  • size: sample size
  • turnout: turnout estimate
  • estimated voting intentions for the main parties

Cleaning the data – surveys

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

# Filter dataset
cleaned_surveys <- surveys |>
  mutate(
    # Parse dates variables as date objects
    field_date_from = ymd(field_date_from),
    field_date_to = ymd(field_date_to),
    date_elec = ymd(date_elec),
    # Calculate the number of fieldwork days
    fieldwork_days = as.numeric(field_date_to - field_date_from + 1)
  ) |>
  filter(
    !exit_poll,                           # Exclude exit polls
    date_elec >= ymd("2008-01-01"),       # Exclude polls referred to elections before 2008
    size >= 750,                          # Exclude polls with sample size < 750
    fieldwork_days > 1                    # Exclude polls with 1 or fewer fieldwork days
  )

# Deleting columns that only have NAs
cleaned_surveys <- cleaned_surveys |> 
  select(where(~ !all(is.na(.))))

# Identify party columns dynamically
metadata_columns <- c("type_survey", "date_elec", "id_pollster", "pollster", "media",
                      "field_date_from", "field_date_to", "fieldwork_days", "exit_poll", 
                      "size", "turnout")
party_columns <- setdiff(colnames(cleaned_surveys), metadata_columns)

# Reshape data into long format
tidy_surveys <- cleaned_surveys |>
  pivot_longer(
    cols = all_of(party_columns),  # Reshape party columns
    names_to = "party_raw",        # Raw party names
    values_to = "votes"            # Corresponding voting intentions
  )

# add on party names by code
tidy_surveys <- tidy_surveys %>%
  mutate(
    party = case_when(
      party_raw == "PSOE" ~ "PARTIDO SOCIALISTA OBRERO ESPAÑOL",
      party_raw == "CIU" ~ "CONVERGÈNCIA I UNIÓ",
      party_raw == "EAJ-PNV" ~ "PARTIDO NACIONALISTA VASCO",
      party_raw == "ERC" ~ "ESQUERRA REPUBLICANA DE CATALUNYA",
      party_raw == "IU" ~ "UNIDAS PODEMOS - IU",
      party_raw == "PP" ~ "PARTIDO POPULAR",
      party_raw == "BNG" ~ "BLOQUE NACIONALISTA GALLEGO",
      party_raw == "CS" ~ "CIUDADANOS",
      party_raw == "EH-BILDU" ~ "EH - BILDU",
      party_raw == "PODEMOS" ~ "UNIDAS PODEMOS - IU",
      party_raw == "VOX" ~ "VOX",
      party_raw == "MP" ~ "MÁS PAÍS",
      TRUE ~ "OTHER")
  )

# Create a column for proper, unqique acronyms
tidy_surveys <- tidy_surveys |> 
  mutate(
    party_code = case_when(
      party == "UNIDAS PODEMOS - IU"~ "PODEMOS-IU",
      party == "OTHER"~ "OTHER",
      TRUE ~ party_raw)
  )

# Select relevant columns
# Getting rid of type_survey, exit_poll (take only 1 value), party_raw
final_surveys <- tidy_surveys |>
  select(-type_survey, -exit_poll, -party_raw) |> 
  relocate(fieldwork_days, .after = field_date_to) |> 
  relocate(votes, .after = party_code) 

# Summing all votes based on the party reclassification
final_surveys <- final_surveys |> 
  group_by(across(-votes)) |> 
  summarize(votes = sum(votes, na.rm = TRUE), .groups = "drop") |> 
  arrange(field_date_from)
# We have 1614 surveys (rows from cleaned_surveys), 12 parties (meaning 12 rows per survey). Thus 1614x12=19368 rows

# Preview
final_surveys

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# Filter dataset
cleaned_surveys <- surveys |>
  mutate(
    # Parse dates variables as date objects
    field_date_from = ymd(field_date_from),
    field_date_to = ymd(field_date_to),
    date_elec = ymd(date_elec),
    # Calculate the number of fieldwork days
    fieldwork_days = as.numeric(field_date_to - field_date_from + 1)
  ) |>
  filter(
    !exit_poll,                           # Exclude exit polls
    date_elec >= ymd("2008-01-01"),       # Exclude polls referred to elections before 2008
    size >= 750,                          # Exclude polls with sample size < 750
    fieldwork_days > 1                    # Exclude polls with 1 or fewer fieldwork days
  )

# Deleting columns that only have NAs
cleaned_surveys <- cleaned_surveys |> 
  select(where(~ !all(is.na(.))))

# Identify party columns dynamically
metadata_columns <- c("type_survey", "date_elec", "id_pollster", "pollster", "media",
                      "field_date_from", "field_date_to", "fieldwork_days", "exit_poll", 
                      "size", "turnout")
party_columns <- setdiff(colnames(cleaned_surveys), metadata_columns)

# Reshape data into long format
tidy_surveys <- cleaned_surveys |>
  pivot_longer(
    cols = all_of(party_columns),  # Reshape party columns
    names_to = "party_raw",        # Raw party names
    values_to = "votes"            # Corresponding voting intentions
  )

# add on party names by code
tidy_surveys <- tidy_surveys %>%
  mutate(
    party = case_when(
      party_raw == "PSOE" ~ "PARTIDO SOCIALISTA OBRERO ESPAÑOL",
      party_raw == "CIU" ~ "CONVERGÈNCIA I UNIÓ",
      party_raw == "EAJ-PNV" ~ "PARTIDO NACIONALISTA VASCO",
      party_raw == "ERC" ~ "ESQUERRA REPUBLICANA DE CATALUNYA",
      party_raw == "IU" ~ "UNIDAS PODEMOS - IU",
      party_raw == "PP" ~ "PARTIDO POPULAR",
      party_raw == "BNG" ~ "BLOQUE NACIONALISTA GALLEGO",
      party_raw == "CS" ~ "CIUDADANOS",
      party_raw == "EH-BILDU" ~ "EH - BILDU",
      party_raw == "PODEMOS" ~ "UNIDAS PODEMOS - IU",
      party_raw == "VOX" ~ "VOX",
      party_raw == "MP" ~ "MÁS PAÍS",
      TRUE ~ "OTHER")
  )

# Create a column for proper, unqique acronyms
tidy_surveys <- tidy_surveys |> 
  mutate(
    party_code = case_when(
      party == "UNIDAS PODEMOS - IU"~ "PODEMOS-IU",
      party == "OTHER"~ "OTHER",
      TRUE ~ party_raw)
  )

# Select relevant columns
# Getting rid of type_survey, exit_poll (take only 1 value), party_raw
final_surveys <- tidy_surveys |>
  select(-type_survey, -exit_poll, -party_raw) |> 
  relocate(fieldwork_days, .after = field_date_to) |> 
  relocate(votes, .after = party_code) 

# Summing all votes based on the party reclassification
final_surveys <- final_surveys |> 
  group_by(across(-votes)) |> 
  summarize(votes = sum(votes, na.rm = TRUE), .groups = "drop") |> 
  arrange(field_date_from)
# We have 1614 surveys (rows from cleaned_surveys), 12 parties (meaning 12 rows per survey). Thus 1614x12=19368 rows

# Preview
final_surveys

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# A tibble: 19,368 × 12
   date_elec  id_pollster pollster media field_date_from field_date_to
   <date>     <chr>       <chr>    <chr> <date>          <date>       
 1 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
 2 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
 3 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
 4 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
 5 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
 6 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
 7 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
 8 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
 9 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
10 2008-03-09 pollster-6  GALLUP   <NA>  2004-04-01      2004-04-21   
# ℹ 19,358 more rows
# ℹ 6 more variables: fieldwork_days <dbl>, size <dbl>, turnout <dbl>,
#   party <chr>, party_code <chr>, votes <dbl>

Creating table for party codes

Creating a table to link each party name to its unique code

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

party_info <- final_surveys |> 
  select(party, party_code) |> 
  unique()

=======

party_info <- final_surveys |> 
  select(party, party_code) |> 
  unique()

>>>>>>> Stashed changes =======

party_info <- final_surveys |> 
  select(party, party_code) |> 
  unique()

>>>>>>> Stashed changes =======

party_info <- final_surveys |> 
  select(party, party_code) |> 
  unique()

>>>>>>> Stashed changes

Cleaning the data – election_data

The election_data file is large and requires quite extensive cleaning to make it “tidy”. We will tidy the data to try make it most useful for future analysis. The election data starts off with 48,737 rows and 471 columns. Reducing the number of columns is a clear priority.

First, we look at the quality of the data and see if any information is redundant and can be removed.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

plot_intro(election_data)

# We see 1.9% missing colums, identify the cols with no data - we have 9 cols. 
blank_cols <- names(election_data)[sapply(election_data, function(x) all(is.na(x)))]

# Drop these columns and also filter to ensure no info outside 2008 to 2019 is included. 
election_data <- election_data |> 
  select(-all_of(blank_cols)) |> 
  filter(anno >= 2008 & anno <= 2019)

# Drop columns that are logical
election_data <- election_data %>%
  select(where(~ !is.logical(.)))

# See the improvements
plot_intro(election_data)

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

plot_intro(election_data)

# We see 1.9% missing colums, identify the cols with no data - we have 9 cols. 
blank_cols <- names(election_data)[sapply(election_data, function(x) all(is.na(x)))]

# Drop these columns and also filter to ensure no info outside 2008 to 2019 is included. 
election_data <- election_data |> 
  select(-all_of(blank_cols)) |> 
  filter(anno >= 2008 & anno <= 2019)

# Drop columns that are logical
election_data <- election_data %>%
  select(where(~ !is.logical(.)))

# See the improvements
plot_intro(election_data)

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

Second, we begin to make the election data tidy. We start by pivoting the data so all columns for party names are within one “party” variable. Before this we have 414 columns referring to parties.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

# Pivot all the party names and ballot counts to the main table
election_pivot <- election_data |> 
  pivot_longer(
    cols = `BERDEAK-LOS VERDES`:`COALICIÓN POR MELILLA`, # select all party data
    names_to = "party",
    values_to = "ballots"
  )
str(election_pivot)

tibble [20,177,118 × 17] (S3: tbl_df/tbl/data.frame)
 $ tipo_eleccion            : chr [1:20177118] "02" "02" "02" "02" ...
 $ anno                     : num [1:20177118] 2008 2008 2008 2008 2008 ...
 $ mes                      : chr [1:20177118] "03" "03" "03" "03" ...
 $ vuelta                   : num [1:20177118] 1 1 1 1 1 1 1 1 1 1 ...
 $ codigo_ccaa              : chr [1:20177118] "14" "14" "14" "14" ...
 $ codigo_provincia         : chr [1:20177118] "01" "01" "01" "01" ...
 $ codigo_municipio         : chr [1:20177118] "001" "001" "001" "001" ...
 $ codigo_distrito_electoral: num [1:20177118] 0 0 0 0 0 0 0 0 0 0 ...
 $ numero_mesas             : num [1:20177118] 2 2 2 2 2 2 2 2 2 2 ...
 $ censo                    : num [1:20177118] 1838 1838 1838 1838 1838 ...
 $ participacion_1          : num [1:20177118] 677 677 677 677 677 677 677 677 677 677 ...
 $ participacion_2          : num [1:20177118] 1008 1008 1008 1008 1008 ...
 $ votos_blancos            : num [1:20177118] 23 23 23 23 23 23 23 23 23 23 ...
 $ votos_nulos              : num [1:20177118] 13 13 13 13 13 13 13 13 13 13 ...
 $ votos_candidaturas       : num [1:20177118] 1269 1269 1269 1269 1269 ...
 $ party                    : chr [1:20177118] "BERDEAK-LOS VERDES" "ARALAR" "PARTIDO OBRERO SOCIALISTA INTERNACIONALISTA" "ALTERNATIVA MOTOR Y DEPORTES" ...
 $ ballots                  : num [1:20177118] 9 27 1 1 2 238 61 85 4 17 ...

head(election_pivot)

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# Pivot all the party names and ballot counts to the main table
election_data <- election_data %>%
  mutate(across(`BERDEAK-LOS VERDES`:`COALICIÓN POR MELILLA`, as.numeric))

Warning: There were 6 warnings in `mutate()`.
The first warning was:
ℹ In argument: `across(`BERDEAK-LOS VERDES`:`COALICIÓN POR MELILLA`,
  as.numeric)`.
Caused by warning:
! NA introdotti per coercizione
ℹ Run `dplyr::last_dplyr_warnings()` to see the 5 remaining warnings.

election_pivot <- election_data |> 
  pivot_longer(
    cols = `BERDEAK-LOS VERDES`:`COALICIÓN POR MELILLA`, # select all party data
    names_to = "party",
    values_to = "ballots"
  )
str(election_pivot)

tibble [21,404,526 × 17] (S3: tbl_df/tbl/data.frame)
 $ tipo_eleccion            : chr [1:21404526] "02" "02" "02" "02" ...
 $ anno                     : num [1:21404526] 2008 2008 2008 2008 2008 ...
 $ mes                      : chr [1:21404526] "03" "03" "03" "03" ...
 $ vuelta                   : num [1:21404526] 1 1 1 1 1 1 1 1 1 1 ...
 $ codigo_ccaa              : chr [1:21404526] "14" "14" "14" "14" ...
 $ codigo_provincia         : chr [1:21404526] "01" "01" "01" "01" ...
 $ codigo_municipio         : chr [1:21404526] "001" "001" "001" "001" ...
 $ codigo_distrito_electoral: num [1:21404526] 0 0 0 0 0 0 0 0 0 0 ...
 $ numero_mesas             : num [1:21404526] 2 2 2 2 2 2 2 2 2 2 ...
 $ censo                    : num [1:21404526] 1838 1838 1838 1838 1838 ...
 $ participacion_1          : num [1:21404526] 677 677 677 677 677 677 677 677 677 677 ...
 $ participacion_2          : num [1:21404526] 1008 1008 1008 1008 1008 ...
 $ votos_blancos            : num [1:21404526] 23 23 23 23 23 23 23 23 23 23 ...
 $ votos_nulos              : num [1:21404526] 13 13 13 13 13 13 13 13 13 13 ...
 $ votos_candidaturas       : num [1:21404526] 1269 1269 1269 1269 1269 ...
 $ party                    : chr [1:21404526] "BERDEAK-LOS VERDES" "ARALAR" "PARTIDO OBRERO SOCIALISTA INTERNACIONALISTA" "ALTERNATIVA MOTOR Y DEPORTES" ...
 $ ballots                  : num [1:21404526] 9 27 1 1 2 238 61 85 4 17 ...

head(election_pivot)

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# A tibble: 6 × 17
  tipo_eleccion  anno mes   vuelta codigo_ccaa codigo_provincia codigo_municipio
  <chr>         <dbl> <chr>  <dbl> <chr>       <chr>            <chr>           
1 02             2008 03         1 14          01               001             
2 02             2008 03         1 14          01               001             
3 02             2008 03         1 14          01               001             
4 02             2008 03         1 14          01               001             
5 02             2008 03         1 14          01               001             
6 02             2008 03         1 14          01               001             
# ℹ 10 more variables: codigo_distrito_electoral <dbl>, numero_mesas <dbl>,
#   censo <dbl>, participacion_1 <dbl>, participacion_2 <dbl>,
#   votos_blancos <dbl>, votos_nulos <dbl>, votos_candidaturas <dbl>,
#   party <chr>, ballots <dbl>

We now have a table with 20,177,118 rows and 17 columns.

This is more clean than previously, but we still need to aggregate of our party variables into the main party groups. We will do this by creating a mapping table (party_names) that standardizes the raw party names into main party groupings (party_main) using regular expressions.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

party_names <- tibble(names = unique(election_pivot$party))

# Party names in the election_data file do not match up perfectly with the abbrev file (i.e. some of the names present in party_names are not in abbrev)
# So it is better to work directly on party_names instead of using abbrev

party_names <- party_names |> 
    mutate(party_main = case_when(
                str_detect(names, "(?i)PSOE|PARTIDO DOS SOCIALISTAS DE GALICIA|PARTIDO SOCIALISTA DE EUSKADI|PARTIDO SOCIALISTA OBRERO ESPAÑOL|PARTIT SOCIALISTA OBRER ESPANYOL") ~ "PARTIDO SOCIALISTA OBRERO ESPAÑOL",
                str_detect(names, "(?i)PARTIDO POPULAR") ~ "PARTIDO POPULAR",
                str_detect(names, "(?i)CIUDADANOS-PARTIDO DE LA CIUDADANIA|CIUDADANOS-PARTIDO DE LA CIUDADANÍA|CIUDADANOS PARTIDO DE LA CIUDADANIA|CIUDADANOS PARTIDO DE LA CIUDADANÍA|CIUDADANOS, PARTIDO DE LA CIUDADANÍA|CIUTADANS") ~ "CIUDADANOS",
                str_detect(names, "(?i)EUZKO ALDERDI JELTZALEA-PARTIDO NACIONALISTA VASCO") ~ "PARTIDO NACIONALISTA VASCO",
                str_detect(names, "(?i)BLOQUE NACIONALISTA GALEGO|BNG") ~ "BLOQUE NACIONALISTA GALLEGO",
                str_detect(names, "(?i)CONVERGENCIA I UNIO|CONVERGÈNCIA I UNIÓ") ~ "CONVERGÈNCIA I UNIÓ",
                str_detect(names, "(?i)PODEM|EZKER BATUA|EZKER ANITZA|IZQUIERDA UNIDA|ESQUERRA UNIDA|ESQUERDA UNIDA") ~ "UNIDAS PODEMOS - IU",
                str_detect(names, "(?i)ESQUERRA REPUBLICANA") ~ "ESQUERRA REPUBLICANA DE CATALUNYA",
                str_detect(names, "(?i)BILDU|EUSKO ALKARTASUNA|ARALAR|SORTU|ALTERNATIBA") ~ "EH - BILDU",
                str_detect(names, "(?i)MÁS PAÍS") ~ "MÁS PAÍS",
                str_detect(names, "(?i)VOX") ~ "VOX",
                TRUE ~ "OTHER")
    )

unique(party_names$party_main)

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

party_names <- tibble(names = unique(election_pivot$party))

# Party names in the election_data file do not match up perfectly with the abbrev file (i.e. some of the names present in party_names are not in abbrev)
# So it is better to work directly on party_names instead of using abbrev

party_names <- party_names |> 
    mutate(party_main = case_when(
                str_detect(names, "(?i)PSOE|PARTIDO DOS SOCIALISTAS DE GALICIA|PARTIDO SOCIALISTA DE EUSKADI|PARTIDO SOCIALISTA OBRERO ESPAÑOL|PARTIT SOCIALISTA OBRER ESPANYOL") ~ "PARTIDO SOCIALISTA OBRERO ESPAÑOL",
                str_detect(names, "(?i)PARTIDO POPULAR") ~ "PARTIDO POPULAR",
                str_detect(names, "(?i)CIUDADANOS-PARTIDO DE LA CIUDADANIA|CIUDADANOS-PARTIDO DE LA CIUDADANÍA|CIUDADANOS PARTIDO DE LA CIUDADANIA|CIUDADANOS PARTIDO DE LA CIUDADANÍA|CIUDADANOS, PARTIDO DE LA CIUDADANÍA|CIUTADANS") ~ "CIUDADANOS",
                str_detect(names, "(?i)EUZKO ALDERDI JELTZALEA-PARTIDO NACIONALISTA VASCO") ~ "PARTIDO NACIONALISTA VASCO",
                str_detect(names, "(?i)BLOQUE NACIONALISTA GALEGO|BNG") ~ "BLOQUE NACIONALISTA GALLEGO",
                str_detect(names, "(?i)CONVERGENCIA I UNIO|CONVERGÈNCIA I UNIÓ") ~ "CONVERGÈNCIA I UNIÓ",
                str_detect(names, "(?i)PODEM|EZKER BATUA|EZKER ANITZA|IZQUIERDA UNIDA|ESQUERRA UNIDA|ESQUERDA UNIDA") ~ "UNIDAS PODEMOS - IU",
                str_detect(names, "(?i)ESQUERRA REPUBLICANA") ~ "ESQUERRA REPUBLICANA DE CATALUNYA",
                str_detect(names, "(?i)BILDU|EUSKO ALKARTASUNA|ARALAR|SORTU|ALTERNATIBA") ~ "EH - BILDU",
                str_detect(names, "(?i)MÁS PAÍS") ~ "MÁS PAÍS",
                str_detect(names, "(?i)VOX") ~ "VOX",
                TRUE ~ "OTHER")
    )

unique(party_names$party_main)

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

 [1] "OTHER"                             "EH - BILDU"                       
 [3] "PARTIDO POPULAR"                   "UNIDAS PODEMOS - IU"              
 [5] "PARTIDO SOCIALISTA OBRERO ESPAÑOL" "PARTIDO NACIONALISTA VASCO"       
 [7] "CIUDADANOS"                        "ESQUERRA REPUBLICANA DE CATALUNYA"
 [9] "CONVERGÈNCIA I UNIÓ"               "BLOQUE NACIONALISTA GALLEGO"      
[11] "VOX"                               "MÁS PAÍS"                         

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

# Adding party code to party_names dataframe
party_names <- party_names |> 
  left_join(party_info, by = join_by(party_main == party))

Now join on the main party names and codes to our election table. Testing was undertaken and the join of a table was more efficient than alternatives (e.g. str_detects over election_pivot or rowwise summaries).

# Join party main and party code into main df
election_pivot <- election_pivot |> 
  left_join(party_names, by = join_by(party == names))

Now we will include some additional information that will make the analysis potentially easier later, including province and total votes counts from our data:

# Create municipal code to join on municipal names. 
# Create total votes column
election_pivot <- election_pivot |>
  mutate(cod_mun = paste(codigo_ccaa, codigo_provincia, codigo_municipio, sep="-"),
         total_votes = votos_blancos + votos_nulos + votos_candidaturas)

# Join municipality names
election_pivot <- election_pivot |> 
  left_join(cod_mun, by = join_by(cod_mun))  

# Check quality of the join and whether NA's have been introduced as municipality names
any(is.na(election_pivot$municipio))

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# Adding party code to party_names dataframe
party_names <- party_names |> 
  left_join(party_info, by = join_by(party_main == party))

Now join on the main party names and codes to our election table. Testing was undertaken and the join of a table was more efficient than alternatives (e.g. str_detects over election_pivot or rowwise summaries).

# Join party main and party code into main df
election_pivot <- election_pivot |> 
  left_join(party_names, by = join_by(party == names))

Now we will include some additional information that will make the analysis potentially easier later, including province and total votes counts from our data:

# Create municipal code to join on municipal names. 
# Create total votes column
election_pivot <- election_pivot |>
  mutate(cod_mun = paste(codigo_ccaa, codigo_provincia, codigo_municipio, sep="-"),
         total_votes = votos_blancos + votos_nulos + votos_candidaturas)

# Join municipality names
election_pivot <- election_pivot |> 
  left_join(cod_mun, by = join_by(cod_mun))  

# Check quality of the join and whether NA's have been introduced as municipality names
any(is.na(election_pivot$municipio))

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

[1] FALSE

Be careful not all 8135 municipalities appear in each election. We have 6 elections and 414 parties, thus we should have 6x414=2484 occurrences for each municipality, but that is not the case.

Also be careful some municipalities have the same name (but different mun_code), so if you ever need to group by municipality remember to group by mun_code instead of municipality.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

# Count the number of times each municipaly appears and then get the unique values for that count (not all are 2484) meaning some municipalities are not present in certain elections
election_pivot |> count(cod_mun) |> pull(n) |> unique()

[1] 2484 1656  828 1242 2070

#Number of unique values for cod_mun is different than number of unique values for municipio
n_distinct(cod_mun$cod_mun)

[1] 8135

n_distinct(cod_mun$municipio)

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# Count the number of times each municipaly appears and then get the unique values for that count (not all are 2484) meaning some municipalities are not present in certain elections
election_pivot |> count(cod_mun) |> pull(n) |> unique()

[1] 2508 1672  836 1254 2926 3344 2090 3762 4180

#Number of unique values for cod_mun is different than number of unique values for municipio
n_distinct(cod_mun$cod_mun)

[1] 8135

n_distinct(cod_mun$municipio)

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

[1] 8118

Now we need to group together all of the votes for “OTHER” variables and create unique identifiers for each individual election in our dataframes.

Currently we have a table of 22 variables with 20,177,118 rows. We can clean this more.

First, identify the redundant data in our election. We can remove:

tipo_eleccion - because all values = 02. It is not useful vuelta = because all values = 1, it is not useful. geographic variables = we will remove codigo_municipio is included in cod_mun which we joined on from the cod_mun table. We keep the autonomous community and proivnce variables for potential future aggregation and analysis. codigo_distrito_electoral - because every value is zero. It is not useful.

Notably, we have many NA ballot rows and a row for each individual party at each election, where will also try to reduce this when we aggregate the party data with the “party_main” variable created.

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

# To clean the data more, reduce our dataset and rename key variables so everything is more consistent in English
tidy_election <- election_pivot |> 
  select(year = anno, 
         month = mes,
         code_community = codigo_ccaa,
         code_province = codigo_provincia,
         code_municipality = cod_mun,
         municipality = municipio,
         population = censo,
         polling_stations = numero_mesas,
         participation_1 = participacion_1,
         participation_2 = participacion_2,
         blank_votes = votos_blancos,
         null_votes = votos_nulos,
         valid_votes = votos_candidaturas,
         total_votes,
         party_main,
         party_code,
         ballots)

summary(tidy_election)

      year         month           code_community     code_province     
 Min.   :2008   Length:20177118    Length:20177118    Length:20177118   
 1st Qu.:2011   Class :character   Class :character   Class :character  
 Median :2016   Mode  :character   Mode  :character   Mode  :character  
=======
=======
>>>>>>> Stashed changes
=======
>>>>>>> Stashed changes
# To clean the data more, reduce our dataset and rename key variables so everything is more consistent in English
tidy_election <- election_pivot |> 
  select(year = anno, 
         month = mes,
         code_community = codigo_ccaa,
         code_province = codigo_provincia,
         code_municipality = cod_mun,
         municipality = municipio,
         population = censo,
         polling_stations = numero_mesas,
         participation_1 = participacion_1,
         participation_2 = participacion_2,
         blank_votes = votos_blancos,
         null_votes = votos_nulos,
         valid_votes = votos_candidaturas,
         total_votes,
         party_main,
         party_code,
         ballots)

summary(tidy_election)

      year         month           code_community     code_province     
 Min.   :2008   Length:21404526    Length:21404526    Length:21404526   
 1st Qu.:2011   Class :character   Class :character   Class :character  
 Median :2015   Mode  :character   Mode  :character   Mode  :character  
<<<<<<< Updated upstream
<<<<<<< Updated upstream
>>>>>>> Stashed changes
=======
>>>>>>> Stashed changes
=======
>>>>>>> Stashed changes
 Mean   :2015                                                           
 3rd Qu.:2019                                                           
 Max.   :2019                                                           
                                                                        
<<<<<<< Updated upstream
<<<<<<< Updated upstream
<<<<<<< Updated upstream
 code_municipality  municipality         population      polling_stations  
 Length:20177118    Length:20177118    Min.   :      3   Min.   :   1.000  
 Class :character   Class :character   1st Qu.:    144   1st Qu.:   1.000  
 Mode  :character   Mode  :character   Median :    454   Median :   1.000  
                                       Mean   :   4249   Mean   :   7.261  
                                       3rd Qu.:   1858   3rd Qu.:   3.000  
                                       Max.   :2384269   Max.   :3742.000  
                                                                           
 participation_1   participation_2    blank_votes         null_votes      
 Min.   :      0   Min.   :      0   Min.   :    0.00   Min.   :    0.00  
 1st Qu.:     57   1st Qu.:     86   1st Qu.:    1.00   1st Qu.:    1.00  
 Median :    185   Median :    278   Median :    3.00   Median :    4.00  
 Mean   :   1640   Mean   :   2448   Mean   :   28.71   Mean   :   29.84  
 3rd Qu.:    720   3rd Qu.:   1109   3rd Qu.:   12.00   3rd Qu.:   16.00  
 Max.   :1022073   Max.   :1531231   Max.   :17409.00   Max.   :16527.00  
                                                                          
  valid_votes       total_votes       party_main         party_code       
 Min.   :      1   Min.   :      2   Length:20177118    Length:20177118   
 1st Qu.:    106   1st Qu.:    109   Class :character   Class :character  
 Median :    336   Median :    343   Mode  :character   Mode  :character  
 Mean   :   3025   Mean   :   3084                                        
 3rd Qu.:   1364   3rd Qu.:   1393                                        
=======
=======
>>>>>>> Stashed changes
=======
>>>>>>> Stashed changes
 code_municipality  municipality         population      polling_stations 
 Length:21404526    Length:21404526    Min.   :      3   Min.   :   1.00  
 Class :character   Class :character   1st Qu.:    147   1st Qu.:   1.00  
 Mode  :character   Mode  :character   Median :    468   Median :   1.00  
                                       Mean   :   4449   Mean   :   7.59  
                                       3rd Qu.:   1930   3rd Qu.:   4.00  
                                       Max.   :2384269   Max.   :3742.00  
                                                                          
 participation_1   participation_2    blank_votes         null_votes      
 Min.   :      0   Min.   :      0   Min.   :    0.00   Min.   :    0.00  
 1st Qu.:     58   1st Qu.:     89   1st Qu.:    1.00   1st Qu.:    1.00  
 Median :    192   Median :    288   Median :    3.00   Median :    4.00  
 Mean   :   1718   Mean   :   2568   Mean   :   30.34   Mean   :   30.93  
 3rd Qu.:    750   3rd Qu.:   1159   3rd Qu.:   12.00   3rd Qu.:   17.00  
 Max.   :1022073   Max.   :1531231   Max.   :17409.00   Max.   :16527.00  
                                                                          
  valid_votes       total_votes       party_main         party_code       
 Min.   :      1   Min.   :      2   Length:21404526    Length:21404526   
 1st Qu.:    109   1st Qu.:    111   Class :character   Class :character  
 Median :    348   Median :    356   Mode  :character   Mode  :character  
 Mean   :   3171   Mean   :   3232                                        
 3rd Qu.:   1420   3rd Qu.:   1449                                        
<<<<<<< Updated upstream
<<<<<<< Updated upstream
>>>>>>> Stashed changes
=======
>>>>>>> Stashed changes
=======
>>>>>>> Stashed changes
 Max.   :1847096   Max.   :1872679                                        
                                                                          
    ballots        
 Min.   :     1    
 1st Qu.:     3    
<<<<<<< Updated upstream
<<<<<<< Updated upstream
<<<<<<< Updated upstream
 Median :    15    
 Mean   :   372    
 3rd Qu.:    93    
 Max.   :919701    
 NA's   :19781159  

tidy_election <- tidy_election |> 
  group_by(across(-ballots))|> 
  summarise(party_ballots = sum(ballots, na.rm=TRUE), .groups = "drop")

tidy_election

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes Median : 14 Mean : 386 3rd Qu.: 94 Max. :919701 NA's :20984544

tidy_election <- tidy_election |> 
  group_by(across(-ballots))|> 
  summarise(party_ballots = sum(ballots, na.rm=TRUE), .groups = "drop")

tidy_election

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# A tibble: 584,844 × 17
    year month code_community code_province code_municipality municipality
   <dbl> <chr> <chr>          <chr>         <chr>             <chr>       
 1  2008 03    01             04            01-04-001         Abla        
 2  2008 03    01             04            01-04-001         Abla        
 3  2008 03    01             04            01-04-001         Abla        
 4  2008 03    01             04            01-04-001         Abla        
 5  2008 03    01             04            01-04-001         Abla        
 6  2008 03    01             04            01-04-001         Abla        
 7  2008 03    01             04            01-04-001         Abla        
 8  2008 03    01             04            01-04-001         Abla        
 9  2008 03    01             04            01-04-001         Abla        
10  2008 03    01             04            01-04-001         Abla        
# ℹ 584,834 more rows
# ℹ 11 more variables: population <dbl>, polling_stations <dbl>,
#   participation_1 <dbl>, participation_2 <dbl>, blank_votes <dbl>,
#   null_votes <dbl>, valid_votes <dbl>, total_votes <dbl>, party_main <chr>,
#   party_code <chr>, party_ballots <dbl>

Joining year and month into one variable

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

final_election <- tidy_election |> 
  mutate(date_elec = paste(year, month, "01", sep = "-")) |> 
  relocate(date_elec, .before = year) |> 
  select(-year, -month)

#Adding correct days to match survey dataframe
final_election <- final_election |> 
  mutate(
    date_elec = ymd(case_when(
      date_elec == "2008-03-01" ~ "2008-03-09",
      date_elec == "2011-11-01" ~ "2011-11-20",
      date_elec == "2015-12-01" ~ "2015-12-20",
      date_elec == "2016-06-01" ~ "2016-06-26",
      date_elec == "2019-04-01" ~ "2019-04-28",
      date_elec == "2019-11-01" ~ "2019-11-10"))
  )

str(final_election)

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

final_election <- tidy_election |> 
  mutate(date_elec = glue("{year}-{month}-01")) |> 
  relocate(date_elec, .before = year) |> 
  select(-year, -month)

#Adding correct days to match survey dataframe
final_election <- final_election |> 
  mutate(
    date_elec = ymd(case_when(
      date_elec == "2008-03-01" ~ "2008-03-09",
      date_elec == "2011-11-01" ~ "2011-11-20",
      date_elec == "2015-12-01" ~ "2015-12-20",
      date_elec == "2016-06-01" ~ "2016-06-26",
      date_elec == "2019-04-01" ~ "2019-04-28",
      date_elec == "2019-11-01" ~ "2019-11-10"))
  )

str(final_election)

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

tibble [584,844 × 16] (S3: tbl_df/tbl/data.frame)
 $ date_elec        : Date[1:584844], format: "2008-03-09" "2008-03-09" ...
 $ code_community   : chr [1:584844] "01" "01" "01" "01" ...
 $ code_province    : chr [1:584844] "04" "04" "04" "04" ...
 $ code_municipality: chr [1:584844] "01-04-001" "01-04-001" "01-04-001" "01-04-001" ...
 $ municipality     : chr [1:584844] "Abla" "Abla" "Abla" "Abla" ...
 $ population       : num [1:584844] 1180 1180 1180 1180 1180 1180 1180 1180 1180 1180 ...
 $ polling_stations : num [1:584844] 2 2 2 2 2 2 2 2 2 2 ...
 $ participation_1  : num [1:584844] 524 524 524 524 524 524 524 524 524 524 ...
 $ participation_2  : num [1:584844] 798 798 798 798 798 798 798 798 798 798 ...
 $ blank_votes      : num [1:584844] 1 1 1 1 1 1 1 1 1 1 ...
 $ null_votes       : num [1:584844] 6 6 6 6 6 6 6 6 6 6 ...
 $ valid_votes      : num [1:584844] 941 941 941 941 941 941 941 941 941 941 ...
 $ total_votes      : num [1:584844] 948 948 948 948 948 948 948 948 948 948 ...
 $ party_main       : chr [1:584844] "BLOQUE NACIONALISTA GALLEGO" "CIUDADANOS" "CONVERGÈNCIA I UNIÓ" "EH - BILDU" ...
 $ party_code       : chr [1:584844] "BNG" "CS" "CIU" "EH-BILDU" ...
 $ party_ballots    : num [1:584844] 0 0 0 0 0 0 19 0 382 512 ...

Election identifiers:

• Timing -> date
• Area information -> code_community (autonomous community), code_province, code_municipality, municipality, population
• General election information -> polling_stations, participation_1, participation_2, blank_votes, null_votes, valid_votes, total_votes
• Party votes received -> party_main, party_code, party_ballots

Creating turnout dataframe

Creating a dataframe storing all the turnout data for each municipality in each election in case we need to work just on turnout or other data that does not change by party.

All this info is still included in final_election

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream

turnout <- final_election |> 
  select(
    date_elec, code_community, code_province, code_municipality, municipality,
    population, polling_stations, participation_1, participation_2, 
    blank_votes, null_votes, valid_votes, total_votes
  ) |> 
  distinct()
turnout

======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

turnout <- final_election |> 
  select(
    date_elec, code_community, code_province, code_municipality, municipality,
    population, polling_stations, participation_1, participation_2, 
    blank_votes, null_votes, valid_votes, total_votes
  ) |> 
  distinct()
turnout

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

# A tibble: 48,737 × 13
   date_elec  code_community code_province code_municipality municipality       
   <date>     <chr>          <chr>         <chr>             <chr>              
 1 2008-03-09 01             04            01-04-001         Abla               
 2 2008-03-09 01             04            01-04-002         Abrucena           
 3 2008-03-09 01             04            01-04-003         Adra               
 4 2008-03-09 01             04            01-04-004         Albanchez          
 5 2008-03-09 01             04            01-04-005         Alboloduy          
 6 2008-03-09 01             04            01-04-006         Albox              
 7 2008-03-09 01             04            01-04-007         Alcolea            
 8 2008-03-09 01             04            01-04-008         Alcóntar           
 9 2008-03-09 01             04            01-04-009         Alcudia de Monteag…
10 2008-03-09 01             04            01-04-010         Alhabia            
# ℹ 48,727 more rows
# ℹ 8 more variables: population <dbl>, polling_stations <dbl>,
#   participation_1 <dbl>, participation_2 <dbl>, blank_votes <dbl>,
#   null_votes <dbl>, valid_votes <dbl>, total_votes <dbl>

Recap cleaning

We have 2 primary datasets at this stage, election data and survey data, plus a turnout dataframe which is a subset of the election data. For surveys, the data has been cleaned so each row represents the votes for one party within a specific national poll. For elections, the data has been cleaned so each row represents the number of votes for a party within an election in a specific municipality.

The final_surveys data includes:

• election date, pollster and media information, fieldwork dates
• size of the survey and turnout
• party name, party code
• votes received (for that party in that poll)

The final_election data includes:

• date of the election
• party name, party code (with non-primary parties grouped)
• identifier for autonomous community, province and municipality
• municipality population
• election information such as number of polling stations, votes per session
• ballots received (for that party per election in each municipality)

The turnout data includes:

• information on the number of votes and type of vote (e.g. valid or blank/null) per municipality in each election.

!!!!!!! WORK ON DATAFRAMES final_surveys, final_election, turnout !!!!!!!
!!!!!!! DO NOT OVERWRITE THESE DATAFRAMES, CREATE NEW ONES IF YOU NEED TO MODIFY THEM (ex. surveys_q1 <- final_surveys) !!!!!!!

Mandatory questions

1.Which party was the winner in the municipalities with more than 100,000 habitants (census) in each of the elections?

2. Which party was the second when the first was the PSOE? And when the first was the PP?

<<<<<<< Updated upstream <<<<<<< Updated upstream <<<<<<< Updated upstream ======= ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

Identify the First and Second Parties

ranked_parties <- final_election |> 
  group_by(date_elec, code_municipality) |> 
  arrange(desc(party_ballots)) |> 
  mutate(rank = row_number()) |> 
  filter(rank <= 2) |>  
  ungroup()

Filter when PSOE is first

psoe_first <- ranked_parties |> 
  filter(rank == 1 & party_code == "PSOE") |> 
  left_join(ranked_parties |> filter(rank == 2), by = c("date_elec", "code_municipality")) |> 
  rename(
    first_party = party_code.x,  
    first_votes = party_ballots.x,  
    second_party = party_code.y,  
    second_votes = party_ballots.y,
    population = population.y
  ) |> 
  select(date_elec, code_municipality, population, first_party, first_votes, second_party, second_votes)

Filter when PP is first

pp_first <- ranked_parties %>%
  filter(rank == 1 & party_code == "PP") %>%  
  left_join(ranked_parties %>% filter(rank == 2), by = c("date_elec", "code_municipality")) |> 
  rename(
    first_party = party_code.x,  
    first_votes = party_ballots.x,  
    second_party = party_code.y,  
    second_votes = party_ballots.y,
    population = population.y
  ) |> 
  select(date_elec, code_municipality, population, first_party, first_votes, second_party, second_votes)

Summarized PSOE first

psoe_summary <- psoe_first %>%
  group_by(date_elec, second_party) %>%
  summarise(
    total_votes = sum(second_votes),  
    count_as_second = n(),           
    .groups = "drop"
  ) %>%
  arrange(desc(date_elec)) 
print(psoe_summary)

# A tibble: 34 × 4
   date_elec  second_party total_votes count_as_second
   <date>     <chr>              <dbl>           <int>
 1 2019-11-10 BNG                    0               1
 2 2019-11-10 CS                   712              15
 3 2019-11-10 EAJ-PNV           118530              10
 4 2019-11-10 EH-BILDU            5840               5
 5 2019-11-10 OTHER              56589             106
 6 2019-11-10 PODEMOS-IU        156999             227
 7 2019-11-10 PP               2541076            2504
 8 2019-11-10 VOX               557759             548
 9 2019-04-28 CS               1369730             629
10 2019-04-28 EAJ-PNV            60411               9
# ℹ 24 more rows

psoe_summary_per_year <- psoe_first %>%
  group_by(date_elec) %>%                                
  filter(second_votes == max(second_votes)) %>%          
  select(date_elec, second_party) %>%      
  arrange(date_elec)  |> 
  print()

# A tibble: 6 × 2
# Groups:   date_elec [6]
  date_elec  second_party
  <date>     <chr>       
1 2008-03-09 PP          
2 2011-11-20 PP          
3 2015-12-20 PP          
4 2016-06-26 PP          
5 2019-04-28 PP          
6 2019-11-10 PP          

PP was the second party when PSOE was the winner in all the elections.

Summarized PP first

pp_summary <- pp_first %>%
  group_by(date_elec, second_party) %>%
  summarise(
    total_votes = sum(second_votes),  
    count_as_second = n(),           
    .groups = "drop"
  ) %>%
  arrange(desc(date_elec)) 
print(pp_summary)

# A tibble: 41 × 4
   date_elec  second_party total_votes count_as_second
   <date>     <chr>              <dbl>           <int>
 1 2019-11-10 BNG                  191               4
 2 2019-11-10 CS                   331              39
 3 2019-11-10 EAJ-PNV              571               4
 4 2019-11-10 ERC                    1               1
 5 2019-11-10 OTHER              25395              85
 6 2019-11-10 PODEMOS-IU          3900              42
 7 2019-11-10 PSOE             1055524            2124
 8 2019-11-10 VOX               190718             572
 9 2019-04-28 BNG                  207               2
10 2019-04-28 CS                 44860             395
# ℹ 31 more rows

pp_summary_per_year <- pp_first %>%
  group_by(date_elec) %>%                                
  filter(second_votes == max(second_votes)) %>%          
  select(date_elec, second_party) %>%      
  arrange(date_elec)  |> 
  print()

# A tibble: 6 × 2
# Groups:   date_elec [6]
  date_elec  second_party
  <date>     <chr>       
1 2008-03-09 PSOE        
2 2011-11-20 PSOE        
3 2015-12-20 PODEMOS-IU  
4 2016-06-26 PODEMOS-IU  
5 2019-04-28 PSOE        
6 2019-11-10 PSOE        

PSOE was the second party when PP was the winner in almost every election. PODEMOS-IU was the second party in the elections of 2015 and 2016 after PP.

Vis PSOE first

ggplot(psoe_first, aes(y = reorder(date_elec, desc(date_elec)), 
                     x = second_votes, fill = second_party)) +
  
  geom_bar(stat = "identity", position = "identity", width = 0.7) +
  scale_fill_manual(values = c("PP" = "#0157a1", "PODEMOS-IU" = "#663278", "VOX" = "#5AB531", "CS" = "#EB6109")) +
  scale_x_continuous(labels = scales::comma) + 
  theme_minimal() +
  labs(x="", y="", title= "Second party when the winner is PSOE", subtitle = "total n of votes per election") +
   theme(
    plot.title = element_text(face = "bold", size = 14, hjust = 0),
    plot.subtitle = element_text(face = "italic", size = 12, hjust = 0),
    legend.position = "bottom",
    legend.title=element_blank(),
    axis.text.y = element_text(size = 10),
    axis.text.x = element_text(size = 10),
    axis.title = element_blank(),
    panel.grid.major.y = element_line(color = "gray", linetype = "dashed", size = 0.3)
  ) 

Warning: The `size` argument of `element_line()` is deprecated as of ggplot2 3.4.0.
ℹ Please use the `linewidth` argument instead.

Second most voted party by population when the first id PSOE

# Step 1: Create population categories with ordered factors
psoe_first <- psoe_first %>%
  mutate(
    population_category = factor(
      case_when(
        population < 10000 ~ "Pueblo Pequeño",
        population >= 10000 & population < 50000 ~ "Pueblo Mediano",
        population >= 50000 & population < 100000 ~ "Pueblo Grande",
        population >= 100000 & population < 500000 ~ "Ciudad Pequeña",
        population >= 500000 & population < 1000000 ~ "Gran Ciudad",
        population >= 1000000 ~ "Metrópolis"
      ),
      levels = c("Pueblo Pequeño", "Pueblo Mediano", "Pueblo Grande", 
                 "Ciudad Pequeña", "Gran Ciudad", "Metrópolis")
    )
  )

# Step 2: Loop through elections and create a plot for each election
unique_dates <- unique(psoe_first$date_elec)
plots <- list()

for (date in  unique_dates) {
  # Ensure `date` is treated as a valid Date object
  current_date <- as.Date(date)
  
  # Filter data for the specific election date
  data_filtered <- psoe_first %>%
    filter(date_elec == current_date) %>%
    group_by(population_category, second_party) %>%
    summarise(
      total_votes = sum(second_votes, na.rm = TRUE),
      .groups = "drop"
    )
  
  # Create the plot
  plot <- ggplot(data_filtered, aes(x = population_category, 
                                    y = total_votes, 
                                    fill = second_party)) +
    geom_bar(stat = "identity", position = "dodge", width = 0.7) +
    scale_fill_manual(values = c(
      "PP" = "#0157a1", 
      "PODEMOS-IU" = "#663278", 
      "VOX" = "#5AB531", 
      "CS" = "#EB6109"
    )) +
    scale_y_continuous(labels = scales::comma) + 
    labs(
      title = paste("Second Party by Population for Election on", format(current_date, "%Y-%m-%d")),  # Format the date properly
      x = "Population Category",
      y = "Total Votes"
    ) +
    theme_minimal() +
    theme(
      plot.title = element_text(face = "bold", size = 14, hjust = 0.5),
      axis.text.x = element_text(size = 10, angle = 45, hjust = 1),
      axis.text.y = element_text(size = 10),
      legend.position = "bottom",
      legend.title = element_blank()  # Remove legend title
    )
  
  # Save the plot to the list
  plots[[as.character(current_date)]] <- plot
}

# Step 3: Display all plots (one at a time)
for (date in unique_dates) {
  print(plots[[as.character(as.Date(date))]])
}

Vis PP first

ggplot(pp_first, aes(y = reorder(date_elec, desc(date_elec)), 
                     x = second_votes, fill = second_party)) +
  
  geom_bar(stat = "identity", position = "identity", width = 0.7) +
  scale_fill_manual(values = c("PSOE" = "#f20400", "PODEMOS-IU" = "#663278", "VOX" = "#5AB531", "CS" = "#EB6109")) +
  scale_x_continuous(labels = scales::comma) + 
  theme_minimal() +
  labs(x="", y="", title= "Second party when the winner is PP", subtitle = "total n of votes per election") +
   theme(
    plot.title = element_text(face = "bold", size = 14, hjust = 0),
    plot.subtitle = element_text(face = "italic", size = 12, hjust = 0),
    legend.position = "bottom",
    legend.title=element_blank(),
    axis.text.y = element_text(size = 10),
    axis.text.x = element_text(size = 10),
    axis.title = element_blank(),
    panel.grid.major.y = element_line(color = "gray", linetype = "dashed", size = 0.3)
  ) 

Second most voted party by population when the first id PSOE

# Step 1: Create population categories with ordered factors for PP
pp_first <- pp_first %>%
  mutate(
    population_category = factor(
      case_when(
        population < 10000 ~ "Pueblo Pequeño",
        population >= 10000 & population < 50000 ~ "Pueblo Mediano",
        population >= 50000 & population < 100000 ~ "Pueblo Grande",
        population >= 100000 & population < 500000 ~ "Ciudad Pequeña",
        population >= 500000 & population < 1000000 ~ "Gran Ciudad",
        population >= 1000000 ~ "Metrópolis"
      ),
      levels = c("Pueblo Pequeño", "Pueblo Mediano", "Pueblo Grande", 
                 "Ciudad Pequeña", "Gran Ciudad", "Metrópolis")
    )
  )

# Step 2: Loop through elections and create a plot for each election for PP
unique_dates_pp <- unique(pp_first$date_elec)
plots <- list()

# Create a list to store plots for PP
plots_pp <- list()

for (date in unique_dates_pp) {
  # Ensure `date` is treated as a valid Date object
  current_date <- as.Date(date)
  
  # Filter data for the specific election date
  data_filtered <- pp_first %>%
    filter(date_elec == current_date) %>%
    group_by(population_category, second_party) %>%
    summarise(
      total_votes = sum(second_votes, na.rm = TRUE),
      .groups = "drop"
    )
  
  # Create the plot
  plot <- ggplot(data_filtered, aes(x = population_category, 
                                    y = total_votes, 
                                    fill = second_party)) +
    geom_bar(stat = "identity", position = "dodge", width = 0.7) +
    scale_fill_manual(values = c(
      "PSOE" = "#f20400", 
      "PODEMOS-IU" = "#663278", 
      "VOX" = "#5AB531", 
      "CS" = "#EB6109"
    )) +
    scale_y_continuous(labels = scales::comma) + 
    labs(
      title = paste("Second Party by Population for Election on", format(current_date, "%Y-%m-%d")),  # Format the date properly
      x = "Population Category",
      y = "Total Votes"
    ) +
    theme_minimal() +
    theme(
      plot.title = element_text(face = "bold", size = 14, hjust = 0.5),
      axis.text.x = element_text(size = 10, angle = 45, hjust = 1),
      axis.text.y = element_text(size = 10),
      legend.position = "bottom",
      legend.title = element_blank()  # Remove legend title
    )
  
  # Save the plot to the list
  plots_pp[[as.character(current_date)]] <- plot
}

# Step 3: Display all plots for PP (one at a time)
for (date in unique_dates_pp) {
  print(plots_pp[[as.character(as.Date(date))]])
}

<<<<<<< Updated upstream <<<<<<< Updated upstream >>>>>>> Stashed changes ======= >>>>>>> Stashed changes ======= >>>>>>> Stashed changes

3. Who benefits from low turnout?

4. How to analyze the relationship between census and vote? Is it true that certain parties win in rural areas?

5. How to calibrate the error of the polls (remember that the polls are voting intentions at national level)?

6. Which polling houses got it right the most and which ones deviated the most from the results?

Additional questions

Confirm with the group your original analysis question to avoid clashing ideas :)

SOME IDEAS FOR THE ORIGINAL QUESTIONS TO START?

• Which regions had the most predictable votes (i.e. consistently voted for the same party) and which regions were the most undecided (i.e. had the most variance in there votes across) between the 2008 and 2019 elections?

• Map the outcomes over time - plotly on the results? Isabelle has some interest but tbc if we will Map.

• Can we load in other data? think that would go down well? Maybe predict the next election results based on previous trends of the 5 years and compare to see if the following election followed the trend? Think Javi would be happy with new data.

• Which municipalities voting patterns were most consistent with the national trends?

• Which 2 media/pollster outlets had the most polarized estimates of each election?

7. Jacklyn and Linghan

8. Yijia and Diego

7. Marco, Isabel, Brad